Objectives To (a) assess the association between dietary diversity (DD) score, socioeconomic status (SES) and maternal body mass index (BMI), and (b) the variation of the effects of DD and SES at different points of the conditional distribution of the BMI.

Methods The study used Demographic and Health Surveys round 5 data sets from Ghana, Namibia and Sao Tome and Principe. The outcome variable for the analysis was maternal BMI. The DD score was computed using 24-hour dietary recall data. Quantile regression (QR) was used to examine the relationship between DD and SES, and maternal BMI, adjusting for other covariates. The QR allows the covariate effects to vary across the entire distribution of maternal BMI.

Results Women who consumed an additional unit of DD achieved an increase of 0.245 in BMI for those in the 90th quantile in Ghana. The effect of household wealth increases for individuals across all quantiles of the BMI distribution and in all the 3 countries. A unit change in the household wealth score was associated with an increase of 0.038, 0.052 and 0.065 units increase in BMI for individuals in the 5th quantile in Ghana, Namibia and Sao Tome and Principe, respectively. Also, 0.237, 0.301 and 0.174 units increased for those in the 90th quantile in Ghana, Namibia and Sao Tome and Principe, respectively. Education had a significant positive effect on maternal BMI across all quantiles in Namibia and negative effect at the 5th, 10th and 90th quantiles in Sao Tome and Principe.

Conclusions There is heterogeneity in the effects of DD and SES on maternal BMI. Studies focusing on the effects of diet and socioeconomic determinants on maternal BMI should examine patterns of effects at different points of the conditional distribution of the BMI and not just the average effect.